A virtual object modification method, device, computer equipment, medium and program product

Abstract

Embodiments of the present application disclose a virtual object modification method and device, computer equipment, medium and program product, which convert natural language of a player into parameters for modification, are not only convenient and fast, reduce the difficulty of modification, and improve the flexibility and individuality of the player's self-defined virtual object. Moreover, in order to improve the modification effect, multiple modifications can be performed through multiple rounds of natural language description. Taking the i-th modification as an example, after obtaining the i-th round of modification text sent through the target account, the historical parameters corresponding to the i-1-th round of modification text from the target account are obtained, the historical parameters can describe the modification intention corresponding to the i-1-th modification, so as to determine the target virtual object corresponding to the i-th modification and the target parameters for describing the modified target virtual object according to the i-th round of modification text and the historical parameters, and then update the target virtual object to realize the modification of the target virtual object. Improve the interaction rate and user retention rate of the game.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, computer device, medium, and program product for modifying virtual objects. Background Technology

[0002] With the continuous development of computer technology, games that allow interaction with virtual objects have emerged, and players are demanding more and more flexibility in their interactions with these virtual objects.

[0003] In related technologies, games provide a variety of functions for interacting with virtual objects. Taking the modification of virtual objects as an example, players can modify the values ​​of various basic attributes of virtual objects, that is, modify virtual objects by modifying parameters.

[0004] However, modifying parameters requires players to have a deep understanding of the game, such as understanding various basic attributes before making changes. Moreover, the modification effect is often poor, resulting in low game interaction rate and low user retention rate. Summary of the Invention

[0005] To address the aforementioned technical problems, this application provides a method, apparatus, computer device, medium, and program product for modifying virtual objects, used to improve game interaction rate and user retention rate.

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

[0007] On one hand, embodiments of this application provide a method for modifying a virtual object, the method comprising:

[0008] Obtain the i-th round of modified text from the target account, where the i-th round of modified text describes the content corresponding to the i-th modification, and i is an integer greater than 1;

[0009] Obtain historical parameters based on the (i-1)th round of modified text, where the (i-1)th round of modified text comes from the target account, and the historical parameters are parameters used to describe the content corresponding to the (i-1)th modification indicated by the (i-1)th round of modified text;

[0010] Based on the i-th round of modified text and the historical parameters, a target virtual object and target parameters are determined, wherein the target virtual object is the virtual object corresponding to the i-th modification, and the target parameters are parameters used to describe the modified target virtual object;

[0011] The target virtual object is updated based on the target parameters to obtain the modified target virtual object.

[0012] On the other hand, embodiments of this application provide a virtual object modification device, the device comprising: an acquisition unit, a determination unit, and a modification unit;

[0013] The acquisition unit is used to acquire the i-th round of modified text from the target account, wherein the i-th round of modified text is used to describe the content corresponding to the i-th modification, and i is an integer greater than 1;

[0014] The acquisition unit is further configured to acquire historical parameters based on the (i-1)th round of modified text, wherein the (i-1)th round of modified text comes from the target account, and the historical parameters are parameters used to describe the content corresponding to the (i-1)th modification indicated by the (i-1)th round of modified text;

[0015] The determining unit is configured to determine a target virtual object and target parameters based on the i-th round of modified text and the historical parameters, wherein the target virtual object is the virtual object corresponding to the i-th modification, and the target parameters are parameters used to describe the modified target virtual object;

[0016] The modification unit is used to update the target virtual object based on the target parameters to obtain the modified target virtual object.

[0017] Optionally, the determining unit is specifically used for:

[0018] The target virtual object is determined based on the i-th round of modified text and the historical parameters;

[0019] Based on the target virtual object and the i-th round of modified text, the target parameters are obtained by conversion using the language model;

[0020] The device further includes a training unit for:

[0021] Multiple modified text samples are obtained, each of which includes a label. The label is a parameter corresponding to the virtual object described by the modified text sample. The multiple modified text samples are used to describe multiple types of virtual objects, and there are multiple virtual objects of each type.

[0022] Based on the modified text sample and the virtual object described by the modified text sample, the prediction parameters are obtained by conversion through the initial language model;

[0023] Based on the difference between the prediction parameters and the labels of the modified text samples, the model parameters of the initial language model are adjusted to obtain the language model trained by the multiple modified text samples.

[0024] Optionally, the determining unit is specifically used for:

[0025] The target virtual object is determined based on the i-th round of modified text and the historical parameters;

[0026] The language model corresponding to the target virtual object is obtained from multiple language models, and different language models are applicable to different types of virtual objects;

[0027] Based on the target virtual object and the i-th round of modified text, the target parameters are obtained by conversion using the language model corresponding to the target virtual object;

[0028] The device further includes a training unit for:

[0029] Multiple modified text samples are obtained to describe the target virtual object. Each modified text sample includes a label, and the label is a parameter corresponding to the target virtual object described by the modified text sample. Different modified text samples correspond to different parameters.

[0030] Based on the modified text sample and the target virtual object, the prediction parameters are obtained by conversion using the initial language model of the target virtual object;

[0031] Based on the difference between the prediction parameters and the labels of the modified text samples, the model parameters of the initial language model of the target virtual object are adjusted to obtain the language model of the target virtual object trained by the multiple modified text samples.

[0032] Optionally, the determining unit is specifically used for:

[0033] If the first pending modification object is obtained by identifying the modified text in the i-th round, then the second pending modification object described by the historical parameters is obtained; if the first pending modification object and the second pending modification object are the same, then the first pending modification object or the second pending modification object is determined as the target virtual object; if the first pending modification object and the second pending modification object are not the same, then the first pending modification object is determined as the target virtual object.

[0034] If the first pending modification object is not obtained by identifying the modified text in the i-th round, then the second pending modification object described by the historical parameters is obtained, and the second pending modification object is determined as the target virtual object.

[0035] Optionally, the determining unit is specifically used for:

[0036] If the first pending modification object and the second pending modification object are different, then the target old version parameters used to describe the first pending modification object are obtained from the old version parameters obtained from the historical modification. The first pending modification object is a virtual object obtained based on the i-th round of modification text, and the second pending modification object is a virtual object obtained based on the (i-1)-th round of modification text. The historical modification refers to the modification made by the target account before the (i-1)-th modification.

[0037] Based on the i-th round of modified text and the target old version parameters, the target virtual object and the target parameters are determined.

[0038] Optionally, the determining unit is specifically used for:

[0039] Obtain the i-th modification time used to identify the i-th modification, and the historical modification time used to identify each of the historical modifications;

[0040] The target old version parameter is determined based on the time difference between the i-th modification time and the historical modification time. The time difference between the historical modification time corresponding to the target old version parameter and the i-th modification time is less than a time threshold, and the virtual object described by the target old version parameter is the first object to be modified.

[0041] Optionally, the determining unit is specifically used for:

[0042] If the first pending modification object and the second pending modification object are the same, then the modification result parameter is determined according to the i-th round of modification text and the historical parameters. The first pending modification object is a virtual object obtained based on the i-th round of modification text, and the second pending modification object is a virtual object obtained based on the (i-1)-th round of modification text. The modification result parameter is a parameter used to characterize the prediction of the modification target of the first pending modification object based on the target account.

[0043] If the difference between the target parameter and the modified result parameter is greater than or equal to the difference threshold, then the direction of the (i+1)th modification is determined based on the modified result parameter.

[0044] If the difference between the target parameter and the modified result parameter is less than the difference threshold, then the modification unit is specifically used for:

[0045] The target virtual object is updated based on the modified result parameters to obtain the modified target virtual object.

[0046] Optionally, if the target parameter includes multiple attribute pairs, and each attribute pair includes an attribute and its corresponding attribute value, then the modification unit is specifically used for:

[0047] If there is a balance relationship between the first attribute and the second attribute, then the primary attribute and the secondary attribute are determined from the first attribute and the second attribute according to the i-th round of modified text and the historical parameters. The first attribute and the second attribute belong to attributes in different attribute pairs. In the balance relationship, a change in the attribute value of the primary attribute will cause a change in the attribute value of the secondary attribute.

[0048] Based on the primary attribute and the balance relationship, determine the range of suitable values ​​for the secondary attribute;

[0049] If the attribute value of the subordinate attribute exceeds the adaptation value range, the attribute value of the subordinate attribute is adjusted to obtain the adjusted attribute value of the subordinate attribute. The adjusted attribute value of the subordinate attribute and the attribute value of the main attribute satisfy the balance relationship.

[0050] Based on the adjusted attribute values ​​of the subordinate attributes, the target parameters are updated to obtain the updated target parameters;

[0051] The target virtual object is updated according to the updated target parameters to obtain the modified target virtual object.

[0052] Optionally, if the balance relationship is a correlation trend curve, then the modification unit is specifically used for:

[0053] Based on the correlation trend curve and the attribute value of the main attribute, determine the appropriate attribute value of the secondary attribute;

[0054] In response to obtaining a confirmation operation for the adapted attribute value, the adapted attribute value is determined as the adjusted attribute value of the subordinate attribute.

[0055] Optionally, if there are multiple target virtual objects, the modification unit is specifically used for:

[0056] Obtain environmental data to describe the virtual environment, where the virtual environment is the environment in which the target virtual object is located;

[0057] Based on the environmental data, multiple target virtual objects are divided into a set of dominant virtual objects and a set of non-dominant virtual objects. The degree of integration between the dominant virtual objects in the set of dominant virtual objects and the environmental data is greater than or equal to a degree of integration threshold, and the degree of integration between the non-dominant virtual objects in the set of non-dominant virtual objects and the environmental data is less than the degree of integration threshold.

[0058] Obtain a first adjustment coefficient and a second adjustment coefficient, wherein the first adjustment coefficient is greater than the second adjustment coefficient;

[0059] The target parameters are adjusted according to the first adjustment coefficient to obtain the first target parameters. The advantageous virtual objects in the advantageous virtual object set are adjusted according to the first target parameters to obtain the adjusted advantageous virtual object set.

[0060] The target parameter is adjusted according to the second adjustment coefficient to obtain the second target parameter. The non-dominant virtual objects in the non-dominant virtual object set are then adjusted according to the second target parameter to obtain the adjusted non-dominant virtual object set.

[0061] The modified target virtual object is obtained based on the adjusted set of dominant virtual objects and the adjusted set of non-dominant virtual objects.

[0062] Optionally, if the target parameter includes attribute pairs, wherein the attribute pairs include attributes described by text and attribute values ​​corresponding to the attributes, then the modification unit is specifically used for:

[0063] Retrieve the correspondence between attributes described by text and attributes described by parameters;

[0064] Based on the correspondence, the attributes described by text in the target parameters are converted into attributes described by parameters to obtain standard parameters;

[0065] The target virtual object is updated based on the standard parameters to obtain the modified target virtual object.

[0066] On the other hand, embodiments of this application provide a computer device, the computer device including a processor and a memory:

[0067] The memory is used to store computer programs and to transfer the computer programs to the processor;

[0068] The processor is configured to execute the methods described above according to instructions in the computer program.

[0069] On the other hand, embodiments of this application provide a computer-readable storage medium for storing a computer program for performing the methods described above.

[0070] On the other hand, embodiments of this application provide a computer program product including a computer program, which, when run on a computer device, causes the computer device to perform the methods described above.

[0071] As can be seen from the above technical solution, the embodiments of this application provide a method for modifying virtual objects. Players do not need to directly modify parameters, but instead modify the corresponding content through natural language description, and then modify the natural language after converting it into parameters. This is not only convenient and quick, reducing the difficulty of modification, but also improving the flexibility and personalization of players' customized virtual objects. Moreover, in order to improve the effect of modification, multiple modifications can be performed through multiple rounds of natural language description. Taking the i-th modification as an example, since the modification intentions corresponding to multiple modifications may be consistent, the i-th round of modification text may not fully describe the modification intention. Therefore, in order to clarify the modification intention, after obtaining the i-th round of modification text sent through the target account, the historical parameters corresponding to the (i-1)-th round of modification text from the target account are obtained. These historical parameters can describe the modification intention corresponding to the (i-1)-th modification. Thus, based on the i-th round of modification text and the historical parameters, the target virtual object corresponding to the i-th modification and the target parameters used to describe the modified target virtual object are determined. Then, the target virtual object is updated based on the target parameters to realize the modification of the target virtual object.

[0072] Therefore, compared to determining the modification intent of the i-th modification by combining the i-1 round of modification text, the historical parameters corresponding to the i-1 round of modification text occupy less storage space, reducing storage pressure. Moreover, for the normal operation of the target virtual object, the historical parameters are content that should be stored. That is, compared to storing the i-1 round of modification text, the historical parameters are stored for a longer period of time. Thus, even if there is a long time difference between the i-th modification and the i-1th modification, the historical parameters used to describe the modification intent corresponding to the i-1th modification can still be obtained. Therefore, no matter how long the interval between two adjacent modifications is, the modification intent can be further clarified through contextual relationships, improving the accuracy of modification intent recognition, improving the modification effect, and thus improving the game's interaction rate and user retention rate. Attached Figure Description

[0073] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0074] Figure 1 This is a schematic diagram illustrating an application scenario of a virtual object modification method provided in an embodiment of this application;

[0075] Figure 2 A flowchart illustrating a virtual object modification method provided in an embodiment of this application;

[0076] Figure 3This is a schematic diagram illustrating an application scenario for modifying a virtual object, as provided in an embodiment of this application.

[0077] Figure 4 A schematic diagram of a conversion module provided in an embodiment of this application;

[0078] Figure 5 A schematic diagram of an optimization module provided in an embodiment of this application;

[0079] Figure 6 A schematic diagram of an effective module provided in an embodiment of this application;

[0080] Figure 7 This is a schematic diagram of the structure of a virtual object modification device provided in an embodiment of this application;

[0081] Figure 8 This application provides a schematic diagram of the structure of a server according to an embodiment of the present application.

[0082] Figure 9 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Detailed Implementation

[0083] The embodiments of this application will now be described with reference to the accompanying drawings.

[0084] Basic attributes are the most fundamental attributes of a virtual object that directly reflect its status. Taking a virtual character as an example, basic attributes can be character attributes such as constitution, strength, intelligence, endurance, and agility. After understanding each basic attribute, players can modify its value. For example, to make the virtual character more friendly, players can increase intelligence and endurance.

[0085] However, basic attributes are customizable options, which not only limits players' imagination but also means multiple players might generate the same virtual object, leading to a homogenized gaming experience, poor modification effects, and reduced game interaction and user retention. Furthermore, the modification effect of a virtual object often involves a combination of multiple attribute values. For example, if a player wants a brave virtual character, they might need to adjust multiple character attributes such as physique, strength, intelligence, and agility, and even secondary attributes like appearance. Modifying parameters requires players to have a deep understanding of each attribute and to repeatedly modify and experiment with various combinations to achieve the desired virtual character. If multiple modifications fail to achieve the desired effect, players may abandon the game, resulting in low interaction and user retention rates.

[0086] Based on this, this application provides a method for modifying virtual objects. Players do not need to directly modify parameters, but instead modify the corresponding content by describing it in natural language. The natural language is then converted into parameters before modification, which is not only convenient and quick, reducing the difficulty of modification, but also improves the flexibility and personalization of players' customized virtual objects. In addition, it can improve the accuracy of recognizing modification intentions in natural language, thereby improving the modification effect and ultimately increasing the game's interaction rate and user retention rate.

[0087] The virtual object modification method provided in this application can be applied to computer devices with virtual object modification capabilities, such as terminal devices and servers.

[0088] Specifically, terminal devices can be desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Smart in-vehicle devices can be in-vehicle navigation terminals and in-vehicle computers, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc., but are not limited to these.

[0089] The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server or server cluster that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. Terminal devices and servers can be connected directly or indirectly via wired or wireless communication; this application does not impose any restrictions on this.

[0090] To facilitate understanding of the virtual object modification method provided in this application embodiment, the following example uses a server as the execution subject of the virtual object modification method to illustrate the application scenarios of the virtual object modification method.

[0091] See Figure 1 This figure is a schematic diagram illustrating an application scenario of a virtual object modification method provided in an embodiment of this application. For example... Figure 1As shown, this application scenario includes a terminal device 110 and a server 120, which can communicate with each other via a communication network. The communication network uses standard communication technologies and / or protocols, typically the Internet, but can also be any network, including but not limited to Bluetooth, a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), mobile, private networks, or any combination of virtual private networks. In some embodiments, customized or dedicated data communication technologies may be used to replace or supplement the aforementioned data communication technologies.

[0092] Terminal device 110 contains a game client for player interaction. This game client provides a game where players can interact with virtual objects, allowing them to modify at least one of these virtual objects. After logging into the game client using a target account on terminal device 110, players can modify the virtual objects multiple times using natural language. For example, in the i-th modification, the player inputs the i-th round of modification text, which is then received by terminal device 110 and sent to server 120.

[0093] Server 120 is the server corresponding to the game client, used to provide virtual object modification services for the game client. Since the modification intents corresponding to multiple modifications may be consistent, the modification text in the i-th round might not fully describe the modification intent. Therefore, to clarify the modification intent, after obtaining the modification text in the i-th round from the target account, server 120 obtains the historical parameters corresponding to the modification text in the (i-1)-th round from the target account. These historical parameters are parameters obtained based on the modification text in the (i-1)-th round, and can describe the modification intent corresponding to the (i-1)-th modification. Thus, based on the modification text in the i-th round and the historical parameters, server 120 determines the target virtual object (modification intent) corresponding to the i-th modification and the target parameters (modification direction) used to describe the modified target virtual object. Then, server 120 updates the target virtual object based on the target parameters, thus modifying the target virtual object, and displays the modified target virtual object through terminal device 110.

[0094] Therefore, players don't need to directly modify parameters. Instead, they can modify the corresponding content by describing it in natural language, and then convert the natural language into parameters before making the changes. This is not only convenient and quick, reducing the difficulty of modification, but also enhances the flexibility and personalization of players' customized virtual objects. Furthermore, it improves the accuracy of recognizing modification intentions in natural language, resulting in better modification effects, and ultimately increasing game interaction and user retention rates.

[0095] The virtual object modification method provided in this application embodiment can be executed by a server. However, in other embodiments of this application, the terminal device may also have similar functions to the server to execute the virtual object modification method provided in this application embodiment, or the terminal device and the server may jointly execute the virtual object modification method provided in this application embodiment. This embodiment does not limit this.

[0096] The following describes in detail a method for modifying virtual objects provided in this application through method embodiments.

[0097] See Figure 2 This figure is a flowchart illustrating a virtual object modification method provided in an embodiment of this application. For ease of description, the following embodiments will still use a server as the execution subject of this virtual object modification method. Figure 2 As shown, the virtual object modification method includes S201-S204.

[0098] S201: Retrieve the i-th round of modified text from the target account.

[0099] The target account is the account that initiates multiple modifications to virtual objects in the game. This application does not specifically limit the target account; it can be the account that logs into the game or an identifier set by the player to identify the initiator of the modification.

[0100] Virtual objects refer to virtual objects in a game, such as virtual environments, virtual characters, and virtual objects. These virtual objects can be virtual characters controlled by the player, teammates of non-player characters (NPCs), third-party characters, or opponents in the game; this application does not specifically limit them.

[0101] The virtual environment is a virtual environment displayed or provided by the client when it runs on the terminal. This virtual environment can be a simulation of the real world, a semi-simulated / semi-fictional three-dimensional world, or a purely fictional three-dimensional world. The virtual environment can be any of a two-dimensional, 2.5-dimensional, or three-dimensional virtual environment. Optionally, the virtual environment is also used for virtual environment battles between at least two virtual characters, and the virtual environment has virtual resources available for use by at least two virtual characters.

[0102] A virtual character refers to an actionable object in a virtual environment. This actionable object can be at least one of a virtual human, a virtual animal, or an anime character. Optionally, when the virtual environment is a three-dimensional virtual environment, the virtual character can be a three-dimensional virtual model, with each virtual character having its own shape and volume within the three-dimensional virtual environment and occupying a portion of the space in the three-dimensional virtual environment. Optionally, the virtual character is a three-dimensional character constructed based on three-dimensional human skeleton technology, and the virtual character achieves different external appearances by wearing different skins. In some implementations, the virtual character can also be implemented using a 2.5D or 2D model; this application does not limit this aspect.

[0103] Players can initiate multiple modifications to virtual objects in the game through a target account. These modifications can be based on the same virtual object or different virtual objects; this application does not specifically limit this. Furthermore, each modification no longer directly alters the attribute values ​​of the virtual object's basic properties, but rather describes the modification intent through natural language. This application does not specifically limit the method of describing the modification intent through natural language, such as issuing a voice message to modify the text, converting the voice message into the modified text, or directly inputting the modified text, etc.

[0104] Taking the i-th modification as an example, where i is an integer greater than 1, the modification intention used to describe the i-th modification is the i-th round of modification text. That is, the i-th round of modification text describes the content corresponding to the i-th modification of the virtual object based on the target account, such as "a frail woman holding an axe," etc. Therefore, by describing the modification intention through natural language, such as the i-th round of modification text, players can freely express their creativity, rather than modifying the attribute values ​​of limited attributes. This greatly enhances the flexibility and personalization of player-customized characters, providing a richer and more player-expected interactive experience.

[0105] All data collected in this application (such as the i-th round of modified text) was collected with the consent and authorization of the data subject (such as a user, organization, or enterprise), and the collection, use, and processing of the relevant data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0106] S202: Obtain historical parameters based on the text modified in the (i-1)th round.

[0107] As mentioned above, to improve the effectiveness of modifications, players may make multiple changes. For example, the text corresponding to the (i-1)th modification might be "a weak woman holding an axe," while the text corresponding to the i-th modification might be "not fierce enough now, make it fiercer." If only the text of the i-th modification is converted, the intention behind the i-th modification might not be correctly identified; only the direction of modification might be obtained, but the object of modification would not be. In other words, multiple modifications may have consistency; for example, the i-th modification might be based on the (i-1)th modification. Therefore, relevant information from previous modifications can be obtained, and the modification intention can be supplemented through contextual understanding, thereby improving the accuracy of identifying the modification intention.

[0108] In related technologies, modified text that has been used is typically saved. For example, after performing the (i-1)th modification, the (i-1)th round of modified text is saved for use in the i-th modification. However, the modified text can be quite large, potentially consuming significant storage space. Therefore, modified text is usually only temporarily stored for a preset duration to reduce storage space usage. For example, with a preset duration of 5 minutes, during the 5 minutes of temporary storage of the (i-1)th round of modified text, even if the i-th round of modified text has not yet been retrieved, the (i-1)th round of modified text will be deleted. Thus, the i-th modification cannot access the modified text used in the (i-1)th round of modified text. In other words, if the time interval between the (i-1)th and i-th modifications is greater than 5 minutes, the (i-1)th round of modified text cannot be retrieved when identifying the modification intent of the i-th modification, potentially affecting the accuracy of identifying the modification intent of the i-th modification.

[0109] Therefore, this application embodiment no longer uses the modified text used in the previous modification to identify the modification intent of the current modification. Instead, it uses the parameters obtained from the previous modification to identify the modification intent of the current modification. Specifically, for each modification text described by the player in natural language, parameters are generated for that round of modification text. Taking the (i-1)th modification as an example, the parameters obtained based on the (i-1)th round of modification text are historical parameters, which are parameters used to describe the content corresponding to the (i-1)th modification indicated by the (i-1)th round of modification text. In other words, since the (i-1)th round of modification text is used to reflect the player's modification intent in the (i-1)th modification, the historical parameters obtained based on the (i-1)th round of modification text can also reflect the player's modification intent.

[0110] Compared to the text modified in round i-1, the historical parameters obtained from the text modified in round i-1 occupy less storage space. Furthermore, after the round i-1 modification, to ensure the virtual object obtained from the round i-1 can function correctly, the historical parameters obtained from the round i-1 modification are generally stored. This allows the virtual object obtained from the round i-1 modification to be derived from the historical parameters. In other words, storing the historical parameters does not incur additional storage space. Therefore, compared to the text modified in round i-1, the historical parameters are stored for a longer period. Even if the last modification was yesterday, a month ago, or even longer ago, the parameters obtained from the previous modification can still be retrieved, thus improving the effectiveness of the current modification. This allows players to more freely use natural language to describe their modification intentions.

[0111] In other words, even if there is a long time difference between the i-th modification and the (i-1)-th modification, it is still possible to obtain historical parameters that describe the modification intent corresponding to the (i-1)-th modification. Thus, no matter how long the interval between two adjacent modifications is, the modification intent can be further clarified through contextual relationships, improving the accuracy of modification intent recognition, improving the modification effect, and thereby improving the game's interaction rate and user retention rate.

[0112] Therefore, after obtaining the modified text of the i-th round from the target account, in order to improve the recognition accuracy, historical parameters obtained from the target account based on the modified text of the (i-1)-th round can be obtained. This allows for subsequent improvement in the accuracy of recognizing the modification intent of the modified text of the i-th round based on the modification intent reflected in the historical parameters. The process of intent recognition is explained below through S203.

[0113] S203: Determine the target virtual object and target parameters based on the modified text and historical parameters in the i-th round.

[0114] Here, the target virtual object is the virtual object corresponding to the i-th modification, that is, the target virtual object is the object to be modified as indicated in the modification intent corresponding to the i-th round of modification text. The target parameter is the parameter used to describe the modified target virtual object, that is, the target parameter is the modification direction indicated in the modification intent corresponding to the i-th round of modification text. Based on the target parameter and the target virtual object, the virtual object corresponding to the i-th modification can be constructed.

[0115] The i-th round of text modification reflects the modification intent corresponding to the i-th modification, while the historical parameters reflect the modification intent corresponding to the (i-1)-th modification, such as the object and direction of the i-th modification. Since the i-th and (i-1)-th modifications are two consecutive modifications from the same account, and multiple modifications are likely to be consistent, the modification intents of the i-th and (i-1)-th modifications are likely to be similar. Therefore, the historical parameters can, to some extent, reflect the modification intent of the i-th modification, thus improving the identification of that intent.

[0116] This application does not specifically limit the method of determining the target virtual object and target parameters. For example, the target virtual object can be determined first based on the i-th round of modified text and historical parameters, and then the target parameters can be obtained based on the target virtual object and the i-th round of modified text. This method will be described in detail later, and will not be repeated here. Another example is that the modification object and modification direction corresponding to the (i-1)-th modification can be determined first based on the historical parameters, and then the target virtual object and target parameters can be obtained based on the i-th round of modified text, the modification object corresponding to the (i-1)-th modification, and the modification direction corresponding to the (i-1)-th modification.

[0117] It should be noted that target parameters generally include the attribute values ​​corresponding to multiple attributes, thus forming the image of a virtual object through these attribute values. Continuing with the example of a "delicate woman holding an axe," the corresponding target parameters could be Constitution = 80, Strength = 90, Intelligence = 60, Endurance = 50, and Agility = 20, with a maximum value of 100 for each attribute. Therefore, through this application, players only need to describe the virtual object they want to obtain the appropriate parameters, without needing to repeatedly adjust the attribute values, making it convenient and quick.

[0118] Furthermore, considering factors such as game balance, multiple attributes may be correlated; for example, a higher attack attribute value may result in a lower defense attribute value. Correlated attributes may derive other attributes; for instance, attack and defense attributes may derive combat efficiency attributes. The balance between attack and defense attributes determines the efficiency and pace of combat. A virtual character with high attack power but low defense may quickly defeat enemies but is also more vulnerable to counterattacks; a virtual character with high defense may excel in prolonged battles. Therefore, through this embodiment, when generating target parameters, not only can the attribute values ​​of basic attributes be generated, but also the attribute values ​​of derived attributes. This makes player modifications more convenient, enriches virtual objects, improves modification effects, and ultimately enhances player interaction efficiency and user retention.

[0119] S204: Update the target virtual object based on the target parameters to obtain the modified target virtual object.

[0120] Before the i-th modification, the parameters of the target virtual object are those obtained from the previous modification. It's understandable that without the previous modification, the parameters of the target virtual object could be 0 or a basic default value. After the i-th modification, the existing parameters of the target virtual object need to be updated, that is, the parameters of the target virtual object are updated to the target parameters, resulting in the modified target virtual object. If the (i+1)-th modification continues to modify the target virtual object, then the parameters of the target virtual object are now the target parameters. These target parameters are then updated to the parameters obtained based on the text modified in the (i+1)-th round, resulting in another modified target virtual object.

[0121] For example, if the text modified in the (i-1)th round is "a weak woman holding an axe" and the text modified in the i-th round is "be fiercer", then the modified target virtual object obtained in the (i-1)th round is a weak woman holding an axe, and the modified target virtual object obtained in the i-th round is a weak woman holding a bigger axe.

[0122] As can be seen from the above technical solution, the embodiments of this application provide a method for modifying virtual objects. Players do not need to directly modify parameters, but instead modify the corresponding content through natural language description, and then modify the natural language after converting it into parameters. This is not only convenient and quick, reducing the difficulty of modification, but also improving the flexibility and personalization of players' customized virtual objects. Moreover, in order to improve the effect of modification, multiple modifications can be performed through multiple rounds of natural language description. Taking the i-th modification as an example, since the modification intentions corresponding to multiple modifications may be consistent, the i-th round of modification text may not fully describe the modification intention. Therefore, in order to clarify the modification intention, after obtaining the i-th round of modification text sent through the target account, the historical parameters corresponding to the (i-1)-th round of modification text from the target account are obtained. These historical parameters can describe the modification intention corresponding to the (i-1)-th modification. Thus, based on the i-th round of modification text and the historical parameters, the target virtual object corresponding to the i-th modification and the target parameters used to describe the modified target virtual object are determined. Then, the target virtual object is updated based on the target parameters to realize the modification of the target virtual object.

[0123] Therefore, compared to determining the modification intent of the i-th modification by combining the i-1 round of modification text, the historical parameters corresponding to the i-1 round of modification text occupy less storage space, reducing storage pressure. Moreover, for the normal operation of the target virtual object, the historical parameters are content that should be stored. That is, compared to storing the i-1 round of modification text, the historical parameters are stored for a longer period of time. Thus, even if there is a long time difference between the i-th modification and the i-1th modification, the historical parameters used to describe the modification intent corresponding to the i-1th modification can still be obtained. Therefore, no matter how long the interval between two adjacent modifications is, the modification intent can be further clarified through contextual relationships, improving the accuracy of modification intent recognition, improving the modification effect, and thus improving the game's interaction rate and user retention rate.

[0124] This application provides two specific implementation methods for S203, namely, the specific implementation method of determining the target virtual object and target parameters based on the i-th round of modified text and historical parameters, which will be described in detail below.

[0125] Method 1: Multiple virtual objects share the same language model.

[0126] Based on the modified text in the i-th round and the historical parameters, the target virtual object is determined; based on the target virtual object and the modified text in the i-th round, the target parameters are obtained by transformation through a language model.

[0127] The language model has contextual understanding capabilities and can be a large language model (LLM), a generative pre-trained transformer (GPT) model, a bidirectional encoder representation from transformers (BERT), etc. This application does not specifically limit it.

[0128] The embodiments of this application do not specifically limit the process of using the language model. For example, the modified text of the i-th round and the target virtual object are input into the language model, and the target parameters are obtained through the language model's transformation. Another example is that a prompt word is constructed based on the modified text of the i-th round and the target virtual object, and the prompt word is input into the language model, and the target parameters are obtained through the language model's transformation, etc.

[0129] Language models inherently possess contextual understanding capabilities, enabling them to comprehend the content of the modified text in the i-th round. Furthermore, they can identify content related to the target virtual object from this modified text and convert it into target parameters. For ease of explanation, the model's ability to convert natural language into parameters is termed its mapping capability. The following section explains this mapping capability through the training process of the language model; see A1-A3 for details.

[0130] A1: Obtain multiple modified text samples.

[0131] Each modified text sample describes a virtual object. Each modified text sample includes a label representing the parameters corresponding to the virtual object it describes. Multiple modified text samples are used to describe various types of virtual objects. For example, "frail woman" and "Grandma Liu" represent different types of virtual objects, while "assassin" and "tank" represent different types of virtual objects. There are multiple virtual objects of each type, and each virtual object has different parameters. Therefore, multiple modified text samples can cover multiple types of virtual objects in at least one game, and each type of virtual object also has multiple versions. This allows the subsequent language model to learn the characteristics of different types of virtual objects, as well as the characteristics of the same virtual object under different parameters.

[0132] A2: Based on the modified text sample and the virtual object described by the modified text sample, the prediction parameters are obtained by transformation through the initial language model.

[0133] Compared to a language model, the initial language model is a model that has not yet been fully trained; that is, the accuracy of the results obtained from the initial language model is lower than the accuracy of the results obtained from the language model. The process of training the initial language model into a language model is explained below.

[0134] Taking one of the multiple modified text samples as an example, based on the modified text sample and the virtual object described by the modified text sample, the initial language model is used to transform and obtain the prediction parameters. The prediction parameters are the parameters used to describe the virtual object described by the modified text sample, which are obtained by the initial language model.

[0135] This application does not specifically limit the method for determining the virtual object described by the modified text sample. For example, the virtual object described by the modified text sample may be obtained simultaneously with the modified text sample. Alternatively, after obtaining the modified text sample, intent recognition may be performed on the modified text sample to obtain the virtual object described by the modified text sample.

[0136] As one possible approach, the initial language model is a pre-trained model, such as BERT or GPT, which is then fine-tuned based on multiple modified text samples to improve the initial language model's understanding of the game context. This, in turn, improves the accuracy of the conversion while reducing the difficulty of model training.

[0137] A3: Based on the difference between the predicted parameters and the labels of the modified text samples, adjust the model parameters of the initial language model to obtain the language model trained with multiple modified text samples.

[0138] The accuracy of the predicted parameters obtained by transforming a modified text sample using the initial language model is low, while the accuracy of the label of the modified text sample is high. Therefore, based on the difference between the predicted parameters and the label, the difference between the predicted parameters and the label is made smaller and smaller. The model parameters of the initial language model are adjusted, and after training with multiple modified text samples, a language model is obtained.

[0139] Through training, the language model can acquire the mapping relationship between modified text samples described in natural language and parameters; that is, the language model learns mapping ability. Moreover, multiple modified text samples not only include various virtual objects, but also multiple instances of each virtual object. Therefore, the language model can not only learn mapping ability but also clearly understand the modification effect of increasing or decreasing each parameter. For example, multiple modified text samples might include phrases like "a weak woman full of optimism and positive energy," "a weak woman full of optimism but not positive energy," "a weak woman who is not optimistic but full of positive energy," and "a weak woman who is generally optimistic and generally positive energy." Through multiple modified text samples, the language model learns the changing trends of the optimism and positive energy attributes, and even the changing trends derived from the coupling of these two attributes.

[0140] Therefore, based on a language model with semantic understanding capabilities, this model is trained to possess mapping capabilities—that is, the ability to convert text samples describing natural language into parameters. Furthermore, based on this mapping capability, it can implicitly learn the changing trends of each attribute, and even the changing trends derived from the coupling of multiple attributes. This allows the language model not only to understand the i-th round of modified text, but also to clearly define the direction of modification to the target virtual object based on this i-th round of modified text. This is similar to the language model matching the i-th round of modified text with multiple modified text samples, fine-tuning based on the best-matching modified text sample, thereby obtaining the target parameters. This improves the accuracy of intent recognition while reducing training difficulty, making it convenient and fast. In addition, this language model has universality, capable of understanding various attributes of multiple virtual objects, as well as the changing trends of various attributes and the changing trends of coupling between attributes, making it applicable to more application scenarios and easier to train.

[0141] Method 2: Different types of virtual objects use different language models.

[0142] Based on the modified text from the i-th round and historical parameters, the target virtual object is determined. In this embodiment, different types of virtual objects use different language models; that is, each virtual object has a corresponding language model. The language model corresponding to the target virtual object is obtained from multiple language models. Based on the target virtual object and the modified text from the i-th round, the target parameters are obtained by converting the language model corresponding to the target virtual object.

[0143] The language model in Method 2 is basically similar to that in Method 1, the only difference being that the language model in Method 1 is universal, applicable to multiple virtual objects, while the language model in Method 2 is specific, applicable only to virtual objects of its corresponding type. The following explanation uses the language model corresponding to the target virtual object type as an example to illustrate the training process of this language model; see B1-B3 for details.

[0144] B1: Obtain multiple modified text samples used to describe the target virtual object.

[0145] Since the language model is trained to work with only one type of virtual object, the modified text samples only need to come from the target virtual object, and there is no need for modified text samples from other types of virtual objects.

[0146] Similarly, each modified text sample includes a label, which represents the parameter of the target virtual object described by the corresponding modified text sample. Different modified text samples have different parameters, meaning different parameters correspond to different target virtual objects. Taking a frail woman as an example, the parameters corresponding to "a frail woman holding an axe" and "a frail woman holding a knife" are different, even though they belong to the same category of target virtual objects.

[0147] B2: Based on the modified text sample and the target virtual object, the prediction parameters are obtained by transforming the initial language model of the target virtual object.

[0148] It should be noted that, compared to step A2, step B2 explicitly states that the virtual object being trained is the target virtual object, and the model used is also the initial language model corresponding to the target virtual object. Other relevant details can be found in step A2.

[0149] B3: Based on the difference between the predicted parameters and the labels of the modified text samples, adjust the model parameters of the initial language model of the target virtual object to obtain the language model of the target virtual object trained through multiple modified text samples.

[0150] It should be noted that, compared to step A3, step B3 explicitly states that the model being trained is the initial language model corresponding to the target virtual object, thereby training a language model suitable for the target virtual object. Other relevant details can be found in step A3.

[0151] Therefore, based on a language model with semantic understanding capabilities, this model is trained to have mapping capabilities—that is, the ability to convert text samples describing natural language into parameters. Furthermore, based on this mapping capability, it can implicitly learn the changing trends of each attribute, and even the changing trends derived from the coupling of multiple attributes. This allows the language model not only to understand the i-th round of modified text, but also to clearly define the direction of modification to the target virtual object based on that i-th round of modified text. This is similar to the language model matching the i-th round of modified text with multiple modified text samples, fine-tuning based on the best-matching modified text sample, and thus obtaining the target parameters. This improves the accuracy of intent recognition while reducing the training difficulty, making it convenient and fast. In addition, this language model is targeted; each type of virtual object corresponds to a language model, thus making the language model more accurate in intent recognition for its applicable virtual objects.

[0152] Furthermore, regardless of whether a general language model or a targeted language model is used for conversion, the problem of excessive player input may arise. For example, the text to be modified in the i-th round might be: "Please draw the image of character XX from book X, and give character XX an axe in their hand. The content of book X is (input the content of book X)." Since book X does not only describe character XX, but also other characters, there are many irrelevant descriptions for character XX. In this case, irrelevant content in the text to be modified in the i-th round can be removed before identifying the modification intent. This application does not specifically limit the removal method; two methods are described below as examples.

[0153] One approach is to construct prompts suitable for the language model. For example, the prompts could include phrases such as "Before determining the target parameters, remove content unrelated to the target virtual object from the i-th round of modified text, and then determine the target parameters," or "Remove non-human descriptions and descriptions related to the historical context from the i-th round of modified text, and then determine the target virtual object and target parameters." This application does not impose any specific limitations on this approach.

[0154] Method Two: Training the language model also enables it to remove irrelevant content. For example, through training, the language model can find relevant information such as background descriptions, appearance descriptions, plot descriptions, and psychological descriptions related to the character from the input content. Based on the i-th round of modified text and the target virtual object, the language model finds relevant information such as background descriptions, appearance descriptions, plot descriptions, and psychological descriptions related to the target virtual object in the i-th round of modified text. Then, based on this relevant information and the i-th round of modified text, the target parameters are obtained. This application does not specifically limit this aspect.

[0155] This application does not specifically limit the method of determining the target virtual object based on the i-th round of modified text and historical parameters. The following describes three cases as examples.

[0156] Scenario 1: The i-th modification does not describe the object being modified.

[0157] Since two consecutive modifications may have consistency, it is possible that the modified object is described in the (i-1)th modification, but not in the ith modification. For example, the text of the (i-1)th round of modifications corresponding to the (i-1)th modification is "a weak woman holding an axe", while the text of the ith round of modifications corresponding to the ith modification is "be more ferocious". Therefore, the text of the ith round of modifications corresponding to the ith modification does not describe the modification.

[0158] The following explanation uses the object modified in the i-th modification as the first object to be modified and the object modified in the (i-1)-th modification as the second object to be modified.

[0159] If the identification of the modified text in the i-th round does not yield the first pending modification object, then the modification object of the i-th modification is likely to be the same as the modification object of the (i-1)-th modification. Thus, the second pending modification object described by the historical parameters obtained in the i-th modification can be obtained. That is, the first pending modification object and the second pending modification object are likely to be the same. Therefore, the second pending modification object is identified as the target virtual object, that is, the modification object targeted by the i-th modification.

[0160] The embodiments of this application do not specifically limit the method of identifying the first object to be modified, such as identifying the i-th round of modified text through a pre-trained recognition model, which can identify the objects included in the text.

[0161] The embodiments of this application do not specifically limit the method of the second object to be modified. For example, the historical parameters may include the object identifier of the second object to be modified, thereby obtaining the second object to be modified based on the object identifier.

[0162] Scenario 2: The i-th modification describes the object being modified, but it is the same as the object being modified in the (i-1)-th modification.

[0163] If the first undetermined modification object is obtained by identifying the modified text in the i-th round, then the second undetermined modification object is obtained based on the historical parameters obtained in the (i-1)-th modification. If the first undetermined modification object and the second undetermined modification object are the same, it means that the modification object of the two modifications is consistent, which meets the characteristic of the consistency of continuous modification. Therefore, the credibility of identifying the first undetermined modification object is high, and the first undetermined modification object or the second undetermined modification object can be determined as the target virtual object.

[0164] Case 3: The i-th modification describes the object being modified, but it is different from the object being modified in the (i-1)-th modification.

[0165] If the first undetermined modification object is obtained by recognizing the modified text in the i-th round, then the second undetermined modification object is obtained based on the historical parameters obtained in the (i-1)-th modification. If the first undetermined modification object and the second undetermined modification object are different, then the modification object of the i-th modification and the modification object of the (i-1)-th modification are inconsistent. This may be due to an abnormality in the recognition of the modified text in the i-th round, or the modification objects of the two modifications are indeed different.

[0166] Based on this, if the reliability of identifying the first object to be modified is high, then it is assumed that the modified objects of the two modifications are truly different, and the first object to be modified can be determined as the target virtual object. If the reliability of identifying the first object to be modified is uncertain, those skilled in the art can set the target virtual object according to their needs, such as determining the first or second object to be modified as the target virtual object. For example, the modified objects corresponding to modifications prior to the (i-1)th modification (such as the (i-2)th modification, the (i-3)th modification, etc.) can be obtained. If the first object to be modified is the same as the modified object of the (i-2)th modification, or the same as the modified object of the (i-3)th modification, then the first object to be modified is determined as the target virtual object.

[0167] Therefore, compared to simply using the object of modification in this modification, determining the object of modification in this modification based on the object of modification in this modification and the object of modification in the previous modification can improve the accuracy of determining the object of modification in this modification, thereby improving the effect of modification on the target virtual object, and thus improving the game's interaction rate and user retention rate.

[0168] As one possible implementation, if the first object to be modified and the second object to be modified are different, although the first object to be modified can be identified as the target virtual object, the historical parameters obtained based on the (i-1)th modification cannot provide relevant support for the target parameters. Therefore, the target parameters can only be determined based on the text modified in the i-th round.

[0169] To further improve the accuracy of the target parameters, the target old-version parameters used to describe the first object to be modified can be obtained from the old-version parameters obtained from previous modifications. Based on the i-th round of modification text and the target old-version parameters, the target virtual object and target parameters are determined. At this point, the target virtual object is the first object to be modified.

[0170] The historical modifications are modifications made by the target account prior to the (i-1)th modification. Each historical modification retrieves the historical modification text, which is the modification text sent by the target account before the (i-1)th modification, such as the (i-2)th round modification text, the (i-3)th round modification text, etc. Each round of modification text is processed using the methods described in S201-S203 above to obtain the parameters corresponding to each round of modification text, i.e., the old version parameters. Each modification yields old version parameters corresponding to a modification object. Multiple modifications can correspond to the same modification object or different modification objects; this application does not specifically limit this.

[0171] Therefore, target old-version parameters describing the first object to be modified can be obtained from the old-version parameters obtained from previous modifications. That is, one or more old-version parameters describing the first object to be modified, i.e., target old-version parameters, can be obtained from the modified objects described by multiple old-version parameters. Although the target old-version parameters and the i-th round of modified text come from two discontinuous modifications, they target the same modified object, and their modification intent may be consistent. Thus, the target parameters are determined based on one or more target old-version parameters describing the first object to be modified.

[0172] It should be noted that the closer the modification round number corresponding to the old version parameter used to describe the first object to be modified is to the i-th round, the greater the likelihood that the modification intention of the old version parameter is closer to the modification intention described in the i-th round modification text. Therefore, the old version parameter obtained from multiple old version parameters describing the first object to be modified, the one closest to the i-th modification, can be determined as the target old version parameter. For example, if the object to be modified in the (i-1)-th modification is different from the object to be modified in the i-th modification, we can determine whether the object to be modified in the (i-2)-th modification is the same as the object to be modified in the i-th modification. If they are the same, the old version parameter obtained based on the (i-2)-th modification is determined as the target old version parameter; if they are different, we can determine whether the object to be modified in the (i-3)-th modification is the same as the object to be modified in the i-th modification. If they are the same, the old version parameter obtained based on the (i-3)-th modification is determined as the target old version parameter; if they are different, we continue to determine whether the object to be modified in the (i-4)-th modification is the same as the object to be modified in the i-th modification, and so on, until the target old version parameter is obtained.

[0173] Therefore, if the objects to be modified in two consecutive modifications are different, such as the first object to be modified being different from the second object to be modified, then from the old parameters corresponding to the previous modifications, the old parameters used to describe the same object as the object to be modified in this modification are selected, namely the target old parameters. The modification intent of the target old parameters may be similar to the modification intent of the i-th round of modification text. Thus, the target old parameters can provide some information that is helpful for this modification, thereby improving the accuracy of the target parameters and improving the game's interaction rate and user retention rate.

[0174] As one possible approach, to avoid the user's modification intention changing due to excessively long intervals between two modifications, one or more old-version parameters describing the first pending modification object can be obtained from multiple old-version parameter descriptions. In the process of obtaining the target old-version parameter, the time difference between the modification time corresponding to the target old-version parameter and the modification time corresponding to the i-th modification is guaranteed to be less than a time threshold, so as to improve the effectiveness of the target old-version parameter in improving the accuracy of the target parameter.

[0175] Specifically, the time of the i-th modification can be obtained to identify the i-th modification, such as by obtaining the time of the i-th round of modified text, etc., and this application does not make specific limitations in this regard. Multiple historical modification times are obtained, each historical modification time being used to identify the modification time of its corresponding historical sub-modification. Here, a historical sub-modification refers to a modification made by the target account before the (i-1)-th modification, i.e., one modification corresponds to one modification time, which can also be the time of obtaining the corresponding modified text, etc.

[0176] It should be noted that, since data related to previous modifications can be deleted (e.g., if 100 modifications have been performed on the target account, data related to the first modification can be deleted), the historical modification time obtained at this time corresponds to the time of the previous modification before deletion, rather than the time corresponding to each individual modification. This reduces storage pressure and computational load.

[0177] After obtaining the i-th modification time and multiple historical modification times, the target old version parameter can be determined based on the time difference between the i-th modification time and the historical modification time. This ensures that the modification object described by the target old version parameter is the same as the first pending modification object, while the time difference between the i-th modification time and the historical modification time corresponding to the target old version parameter is less than the time threshold.

[0178] The embodiments of this application do not specifically limit the size of the time threshold. Those skilled in the art can set it according to time needs, such as a week.

[0179] This application does not specifically limit the method of determining the target old version parameter based on the time difference between the i-th modification time and the historical modification time. For example, it can determine the time difference between the i-th modification time and each historical modification time, obtain the historical modification time corresponding to the time difference less than the time threshold from multiple historical modification times, and obtain the old version parameter used to describe the first object to be modified from the old version parameters corresponding to the historical modification time corresponding to the time difference less than the time threshold.

[0180] Therefore, if the objects to be modified in two consecutive modifications are different, the target old parameter can be selected from the old parameters corresponding to the previous modifications. The target old parameter is the same as the object to be modified in the current modification, and the difference between the modification time and the modification time of the current modification is less than the time threshold. The modification intent of the target old parameter is more likely to be similar to the modification intent of the i-th round of modification text. This avoids the problem of changes in modification tendency caused by too long a time interval. As a result, the target old parameter can provide some information that is helpful for the current modification, thereby improving the accuracy of the target parameter and improving the game's interaction rate and user retention rate.

[0181] Research has shown that while players can describe their intended modifications using natural language, the i-th round of modification text may not accurately depict the player's final desired direction due to ambiguity or other reasons, potentially requiring several more modifications to achieve the desired goal. To reduce the number of modifications required, this application predicts the user's true intended direction of modification, obtains modification result parameters, and updates the target virtual object based on these parameters, as detailed in C1-C3.

[0182] C1: If the first pending modification object and the second pending modification object are the same, then determine the modification result parameters based on the i-th round of modification text and historical parameters.

[0183] The first pending modification object is a virtual object obtained based on the text modified in the i-th round, and the second pending modification object is a virtual object obtained based on the text modified in the (i-1)-th round. If the first and second pending modification objects are the same, it means that the two adjacent modifications are targeting the same object, and the modification intentions of these two modifications are likely to be consistent. Therefore, the modification result parameters can be predicted based on the historical parameters obtained from the text modified in the i-th round and the (i-1)-th modification. The modification result parameters are used to characterize the modification target of the first pending modification object based on the target account. In other words, the modification result parameters are parameters predicted based on the modification intentions of two adjacent modifications, used to characterize the modification intentions that the player actually wants to express, such as the parameters obtained when the modification of the target virtual object was completed in the (i+1)-th or (i+3)-th round.

[0184] The embodiments of this application do not specifically limit the method of determining the modification result parameters based on the modified text in the i-th round and historical parameters. For example, a prediction model for prediction can be trained, and then the modification result parameters can be obtained by prediction based on the target parameters and historical parameters corresponding to the modified text in the i-th round.

[0185] As one possible implementation, to improve the accuracy of predicting modification result parameters, the modification text corresponding to previous modifications can be obtained. From this text, target modification text describing the first object to be modified can be extracted. There can be one or more target modification texts. Based on these target modification texts, the i-th round of modification text, and historical parameters, the modification result parameters can be determined. Furthermore, by using more modification text describing the first object to be modified, the player's modification intentions become clearer, improving the accuracy of predicting the modification result parameters.

[0186] C2: If the difference between the target parameter and the modified result parameter is greater than or equal to the difference threshold, then the direction of the (i+1)th modification is determined based on the modified result parameter.

[0187] If the difference between the target parameter and the modified result parameter is greater than or equal to the difference threshold, it indicates that there is a large difference between the target parameter and the modified result parameter. This means that multiple modifications may be needed to achieve the player's true intention. To reduce the number of modifications required, the direction of the (i+1)th modification can be determined based on the modified result parameter. This direction can provide the player with suggestions for the next modification, thereby helping the player quickly complete the modification that meets their true intention and improving the player's user experience.

[0188] In addition, this modification, i.e. the i-th modification, updates the target virtual object based on the target parameters to obtain the modified target virtual object, and displays the direction of the i+1-th modification to the player.

[0189] C3: If the difference between the target parameter and the modified result parameter is less than the difference threshold, then the target virtual object is updated based on the modified result parameter to obtain the modified target virtual object.

[0190] If the difference between the target parameter and the modified result parameter is less than the difference threshold, it means that the target parameter and the modified result parameter are not significantly different. In order to help players quickly complete the modification that meets their true modification intentions, the modified result parameter can be replaced with the target parameter. That is, the target virtual object is updated based on the modified result parameter to obtain the modified target virtual object, thus eliminating the need for another modification. This reduces the number of modifications while meeting user needs.

[0191] Therefore, by predicting the modification result parameters that characterize the player's true modification intentions, we can gain a deeper understanding of those intentions. If the modification result parameters are not significantly different from the target parameters, it indicates that the target parameters representing the i-th modification are close to the player's true modification intentions. In this case, we can directly modify the target virtual object based on the modification result parameters, reducing the number of modifications the player needs to make. However, if the modification result parameters differ significantly from the target parameters, directly modifying based on the result parameters may lead to a misunderstanding of the player's modification intentions. For example, if the player wants to see the effect of the current modification, we can display the direction of the (i+1)-th modification while modifying based on the target parameters, helping the player quickly achieve the desired effect. This improves the accuracy of the target parameters, thereby increasing the game's interaction rate and user retention rate.

[0192] As one possible implementation, the target parameters include multiple attribute pairs, each pair consisting of an attribute and its corresponding value. It's important to note that these attributes may have a balance relationship. For example, in a balance relationship between two attributes, a change in the value of one attribute will cause a change in the value of the other. For instance, if the value of one attribute increases, the value of its balancing attribute will decrease; similarly, a virtual character with high attack power generally has low defense. Therefore, by establishing a balance relationship between multiple attributes in the game, the game's balance, internal logic, and performance requirements can be ensured.

[0193] However, in practical applications, players may disregard game settings and generate the i-th round of modified text solely based on their expectations of the virtual object. This could lead to target parameters generated based on the i-th round of modified text not conforming to game settings. Therefore, this application's embodiments fine-tune the target parameters to obtain updated target parameters suitable for game settings. The target virtual object is then updated based on these updated target parameters, ensuring that the modified target virtual object meets both player expectations and game settings, thus guaranteeing normal game operation. The following explanation uses the first and second attributes, which have a balanced relationship among the multiple attributes included in the target parameters, as an example; see D1-D5 for details.

[0194] D1: If there is a balance between the first attribute and the second attribute, then determine the primary and secondary attributes from the first and second attributes based on the modified text and history parameters in the i-th round.

[0195] The first attribute is an attribute in one of the multiple attribute pairs included in the target parameter, and the second attribute is an attribute in one of the multiple attribute pairs included in the target parameter. The first attribute and the second attribute are different attributes.

[0196] If there is a balance between the primary and secondary attributes, it means that a change in the value of the primary attribute will also change the value of the secondary attribute, or vice versa. In this case, the primary attribute is more important than the secondary attribute. For example, if there is a balance between attack and defense attributes, and the virtual character is focused on output, then attack is the primary attribute and defense is the secondary attribute.

[0197] The importance of an attribute can be determined based on one or more factors, such as the game's different emphasis on the first and second attributes, or the player's different expectations for modifying the virtual object. For example, based on the text modified in the i-th round and the historical parameters, the primary and secondary attributes can be distinguished from the first and second attributes, that is, the primary attribute and the secondary attribute can be obtained.

[0198] Therefore, if there is a balanced relationship between the first and second attributes, they can be used to describe the i-th round of modification text and history parameters that the player intends to modify, determining the primary and secondary attributes from the first and second attributes. If the first attribute is the primary attribute, then the second attribute is the secondary attribute; if the second attribute is the primary attribute, then the first attribute is the secondary attribute. That is, the primary and secondary attributes are different attributes.

[0199] D2: Based on the primary attribute and the balance relationship, determine the range of suitable values ​​for the secondary attribute.

[0200] Primary and secondary attributes each have their own attribute value ranges. For example, the attribute value range for the attack attribute (i.e., attack power) is 0-100, and the attribute value range for the defense attribute (i.e., defense power) is also 0-100. The target parameters obtained based on the modified text in the i-th round and the historical parameters conform to their respective attribute value ranges, such as attack power being 80 and defense power being 80.

[0201] However, there is a balance between primary and secondary attributes; for example, the sum of attack power and defense power cannot exceed 120. Based on this, the appropriate value range for secondary attributes can be determined based on the balance relationship and the primary attribute. That is, the attribute value range of the secondary attribute that conforms to the balance relationship, such as the secondary attribute value not exceeding 40.

[0202] D3: If the attribute value of the dependent attribute exceeds the adaptive value range, adjust the attribute value of the dependent attribute to obtain the adjusted attribute value.

[0203] If the attribute value of a secondary attribute exceeds the adaptive value range, it means that the current attribute value of the secondary attribute and the attribute value of the primary attribute do not satisfy the balance relationship. In order to ensure the game balance, the attribute value of the secondary attribute can be readjusted based on the balance relationship to obtain the adjusted attribute value of the secondary attribute. The adjusted attribute value of the secondary attribute and the attribute value of the primary attribute satisfy the balance relationship.

[0204] If the attribute value of the secondary attribute does not exceed the adaptive value range, it means that the attribute value of the secondary attribute that the player expects satisfies the balance relationship with the attribute value of the primary attribute. In this case, there is no need to adjust the attribute value of the secondary attribute.

[0205] This application does not specifically limit the method of adjusting the dependent attribute. For example, the attribute value of the dependent attribute can be scaled. Alternatively, a scaling factor can be determined based on the difference between the attribute value of the primary attribute and its corresponding upper limit value, and then the attribute value of the dependent attribute can be scaled based on this scaling factor. Another example is directly adjusting the attribute value of the dependent attribute to its corresponding upper or lower limit value. The following explanation uses two attributes whose balance relationship can be described by a correlation trend curve as an example.

[0206] This correlation trend curve describes the mutual influence between the primary and secondary attributes. If the correlation trend curve is plotted using a Cartesian coordinate system, the x-axis represents the primary attribute value, and the y-axis represents the secondary attribute value. Based on the correlation trend curve and the primary attribute value, a suitable attribute value for the secondary attribute can be determined. This suitable attribute value and the primary attribute value satisfy the correlation trend curve, as shown by a point on the curve. Players can be prompted that the suitable attribute value is more appropriate for the i-th modification. In response to the determination of the suitable attribute value, i.e., if the player uses this suitable attribute value, then this suitable attribute value is determined as the adjusted attribute value of the secondary attribute.

[0207] Therefore, if the balance between primary and secondary attributes can be described based on the correlation trend curve, the appropriate attribute value for the secondary attribute can be determined based on the correlation trend curve and the attribute value of the primary attribute. After the player confirms, the appropriate attribute value is directly set as the adjusted attribute value of the secondary attribute, resulting in the updated target parameter. This ensures that the updated target parameter meets the player's needs without disrupting the game's design or maintaining game balance.

[0208] D4: Based on the adjusted attribute values ​​of the subordinate attributes, update the target parameters to obtain the updated target parameters.

[0209] The target parameters can be obtained by taking each pair of attributes with a balanced relationship as the first and second attributes, respectively, and then obtaining the adjusted attribute value of each attribute, thus obtaining the updated target parameters.

[0210] D5: Update the target virtual object based on the updated target parameters to obtain the modified target virtual object.

[0211] Therefore, if the target parameters include multiple attributes, among which two attributes have a balanced relationship, it's possible to determine which attribute is the primary attribute and which is the secondary attribute. Based on the primary attribute and the balance relationship, the appropriate value range for the secondary attribute is determined. If the player's desired value for the secondary attribute exceeds the appropriate value range, the secondary attribute is adjusted based on the balance relationship to ensure that the secondary attribute and primary attribute satisfy the balance, thus obtaining the updated target parameters. Furthermore, the updated target parameters, including both primary and secondary attributes, not only meet the player's expectations to a certain extent but also maintain a balance, ensuring that the target virtual object conforms to the game settings, thereby guaranteeing the gaming experience for other players.

[0212] In addition, the relationships between attributes can be analyzed. Taking two different attributes, such as the third attribute and the fourth attribute, as an example, the change in the value of the third attribute will cause the change in the value of the fourth attribute. In the balance relationship between the third attribute and the fourth attribute, the third attribute must be the primary attribute. In this case, the third attribute can be determined as the basic attribute and the fourth attribute as the subordinate attribute of the third attribute. When calculating the target parameter, the basic attribute can be determined first based on the modified text in the i-th round and the historical parameters, and then the subordinate attribute can be determined based on the basic attribute to reduce the amount of calculation.

[0213] As one possible implementation, since the target parameter includes multiple attributes, there are multiple ways to combine attribute values. These multiple combinations may lead to different final modification effects. Therefore, in order to find the optimal combination of attribute values, multi-objective optimization algorithms such as genetic algorithms and simulated annealing algorithms can be used to obtain the optimal attribute values ​​for multiple attributes, thereby improving the modification effect.

[0214] As one possible implementation, if multiple target virtual objects are determined based on the modified text and historical parameters in the i-th round, then multiple target virtual objects can be adjusted in batches based on the specific environment or rules of the game, as detailed in E1-E6.

[0215] E1: Obtain environmental data used to describe the virtual environment.

[0216] Environmental data in a virtual environment is used to describe its characteristics, such as weather data that might affect the attribute values ​​(e.g., attack power) of virtual characters in a game, and specific game rules that might affect attribute values. For example, if the game season is a "Rampage Season," then attack-type virtual characters might receive a greater increase in attack power than non-attack-type virtual characters.

[0217] E2: Divide multiple target virtual objects based on environmental data to obtain a set of dominant virtual objects and a set of non-dominant virtual objects.

[0218] The set of advantageous virtual objects includes 0 or more advantageous virtual objects. The degree of integration between advantageous virtual objects and environmental data is greater than or equal to the degree of integration threshold. For example, attack-type virtual characters are more adapted to the characteristics of the Frenzy Season, that is, attack-type virtual characters can exert their advantages more in the Frenzy Season. Therefore, in the Frenzy Season, attack-type virtual characters will be classified into the set of advantageous virtual objects.

[0219] The set of non-dominant virtual objects includes 0 or more non-dominant virtual objects. Non-dominant virtual objects have a lower degree of integration with environmental data than the integration threshold. For example, non-aggressive virtual characters are less likely to exert their advantages in the Frenzy Season. Therefore, in the Frenzy Season, non-aggressive virtual characters will be classified into the set of non-dominant virtual objects.

[0220] The embodiments of this application do not specifically limit the fusion degree threshold, and those skilled in the art can set it according to actual needs.

[0221] E3: Obtain the first adjustment factor and the second adjustment factor.

[0222] The first adjustment coefficient is greater than the second adjustment coefficient. The first adjustment coefficient is used to adjust the attribute values ​​of the dominant virtual objects, and the second adjustment coefficient is used to adjust the attribute values ​​of the non-dominant virtual objects.

[0223] E4: Adjust the target parameters according to the first adjustment coefficient to obtain the first target parameters, and adjust the dominant virtual objects in the dominant virtual object set according to the first target parameters to obtain the adjusted dominant virtual object set.

[0224] If the set of advantageous virtual objects does not include advantageous virtual objects, then there is no need to adjust the set of advantageous virtual objects; if the set of advantageous virtual objects includes one or more advantageous virtual objects, then the advantageous virtual objects in the set of advantageous virtual objects can be adjusted based on the first adjustment coefficient, so that the attribute values ​​of the advantageous virtual objects are increased, so that the advantageous virtual objects can exert a greater advantage in their more suitable virtual environment.

[0225] As one possible approach, the dominant attribute among the multiple attributes of the dominant virtual object can be determined based on environmental data. That is, the attribute that matches the environmental data among the multiple attributes. Then, the dominant attribute is adjusted according to the first adjustment coefficient to obtain the adjusted dominant virtual object, thereby obtaining the set of adjusted dominant virtual objects, and thus achieving targeted adjustment.

[0226] E5: Adjust the target parameters according to the second adjustment coefficient to obtain the second target parameters. Adjust the non-dominant virtual objects in the non-dominant virtual object set according to the second target parameters to obtain the adjusted non-dominant virtual object set.

[0227] If the set of non-dominant virtual objects does not include non-dominant virtual objects, then there is no need to adjust the set of non-dominant virtual objects. If the set of non-dominant virtual objects includes one or more non-dominant virtual objects, then the non-dominant virtual objects in the set of non-dominant virtual objects can be adjusted based on the second adjustment coefficient, so that the attribute values ​​of the non-dominant virtual objects increase, but the increase is small, so that the non-dominant virtual objects do not fail to exert their advantages in their unsuitable virtual environments.

[0228] E6: Based on the adjusted set of dominant virtual objects and the adjusted set of non-dominant virtual objects, obtain the modified target virtual object.

[0229] Therefore, by using the first adjustment coefficient and the second adjustment coefficient, not only can multiple target virtual objects be fine-tuned in batches, but also dominant and non-dominant virtual objects can be determined based on the environmental data of the virtual environment in which the target virtual objects are located. Thus, dominant virtual objects are adjusted based on the larger first adjustment coefficient, and non-dominant virtual objects are adjusted based on the smaller second adjustment coefficient, making the dominant virtual objects more compatible with the virtual environment, thereby highlighting the different effects brought about by the fine-tuning, and making the adjusted target virtual objects more adaptable to the game environment and rules.

[0230] As one possible implementation, to improve scalability and maintainability, the target parameters can be made more readable. For example, attributes can be described using highly readable methods such as text descriptions, or attribute values ​​can be described using a combination of characters and numbers. This allows the target parameters to be displayed before adjustments are made to the virtual object, allowing players to fine-tune them. For instance, the units used in the front-end and back-end might differ; for example, if the front-end shows a 5% increase in intelligence, the back-end might display the specific numerical value corresponding to the increased intelligence attribute.

[0231] Specifically, if the target parameter includes attribute pairs, which include attributes described by text and attribute values ​​corresponding to the attributes, the target parameter is displayed before updating the target virtual object. In response to a modification operation on the target parameter, an update parameter is obtained based on the modification operation and the target parameter, and the target virtual object is updated based on the update parameter.

[0232] However, if the target parameter or update parameter includes text, the backend may have difficulty understanding the text when updating the target virtual object based on the target parameter or update parameter, leading to problems in the update process. The following explanation uses the target parameter as an example.

[0233] Obtain the correspondence between attributes described by text and attributes described by parameters. Based on the correspondence, convert the attributes described by text in the target parameters into attributes described by parameters to obtain standard parameters. Update the target virtual object based on the standard parameters to obtain the modified target virtual object.

[0234] Taking an attack attribute as an example, the target parameter describes the attack attribute as "Attack attribute: 20". In the correspondence between attributes described by text and attributes described by parameters, the attack attribute and 'a' have an object relationship. Therefore, "Attack attribute: 20" can be converted to "a: 20", meaning the standard parameter is "a: 20". The standard parameter only includes characters, making it convenient for the backend to update the target virtual object based on the standard parameter.

[0235] As one possible implementation, the attribute corresponding to each position in the standard parameter can be specified, such as by representing the standard parameter through an array, where the value in the 3rd row and 6th column of the array corresponds to the attribute value corresponding to the attack attribute, thus facilitating the storage of the standard parameter.

[0236] Therefore, although target parameters can be represented in a more readable way, such as text and numbers, rendering directly based on target parameters in this format may lead to problems such as difficulty in understanding and poor modification results. Based on this, target parameters can be converted into a machine-friendly format, such as full parameters, to obtain the converted standard parameters. The target virtual object can then be updated based on the standard parameters to improve the modification effect of the modified target virtual object.

[0237] As one possible approach, to increase personalization and accuracy, a virtual object modification method applicable to each game can be set based on the data of each game. However, a game can also have multiple versions. After generating the target parameters, the target parameters can be batch-fine-tuned based on the different versions applicable to the game, so that the fine-tuned target parameters use their corresponding versions respectively.

[0238] As one possible implementation, since modified texts from multiple accounts may be received simultaneously, resulting in high concurrency, a cluster can be used to address this issue. For example, the cluster could include a first device and a second device. If the first device receives modified texts from multiple accounts, it can forward the modified texts of some of the accounts to the second device so that the second device can process the modified texts of those accounts, thereby modifying the corresponding virtual objects.

[0239] It's worth noting that this cluster can employ a microservice architecture, enabling collaboration between different devices, even different modules within the same device, or different modules across different devices, through interfaces to ensure cluster stability and security. Furthermore, modules within the cluster can be containerized to enable rapid deployment and migration, thereby improving module independence and scalability.

[0240] To facilitate a further understanding of the technical solutions provided in the embodiments of this application, the following description takes the execution subject of the virtual object modification method provided in the embodiments of this application as a server as an example, and provides an overall exemplary introduction to the virtual object modification method.

[0241] In open-world games, there are many customizable objects—virtual objects whose values ​​can be freely defined by players, including but not limited to character customization and adjusting virtual object parameters. Using the virtual object modification method provided in this application, players can describe their modification intentions using natural language and can make multiple modifications.

[0242] See Figure 3 This figure is a schematic diagram illustrating an application scenario for modifying a virtual object according to an embodiment of this application. Figure 3 The system includes an acquisition module 301, a conversion module 302, an optimization module 303, and an activation module 304. Each module will be described below.

[0243] (1) Acquisition module 301 is used to acquire the modified text of the player's current modification and send the modified text to conversion module 302.

[0244] The following explanation uses the i-th modification as an example. The modified text in this modification is the i-th round of modification text from the target account. After obtaining the i-th round of modification text, the historical parameters obtained based on the (i-1)-th round of modification text are retrieved, and the i-th round of modification text and the historical parameters are sent to the conversion module 302.

[0245] (2) The conversion module 302 is used to obtain the historical parameters based on the modified text of the (i-1)th round after obtaining the modified text of the i-th round, determine the target virtual object and the target parameters according to the modified text of the i-th round and the historical parameters, and send the target parameters and the target virtual object to the optimization module 303.

[0246] See Figure 4 This figure is a schematic diagram of a conversion module provided in an embodiment of this application. Figure 4 In the middle, the conversion module 302 includes a natural language processing engine 3021, a context manager 3022, a dialogue state follower 3023, an intent recognizer 3024, and a language model 3025, which will be described below.

[0247] The Natural Language Processing Engine (NLP Engine) 3021 is responsible for performing natural language processing (NLP) on the i-th round of modified text input by the player, which is described in natural language. This includes word segmentation, part-of-speech tagging, and named entity recognition.

[0248] The Context Manager 3022 is used to maintain the dialogue state and follow the evolution of user intent, ensuring that the module can understand the user's coherent intent in multiple rounds of dialogue. For example, it can retrieve historical parameters obtained from the (i-1)th round of modified text based on the i-th round of modified text, or retrieve legacy parameters that are the same as those described in the i-th round of modified text. The retrieval of legacy parameters can be found in the aforementioned related content and will not be repeated here.

[0249] The Dialogue State Tracker 3023 records key information during the dialogue process, providing necessary contextual information for the Intent Recognizer 3024, such as obtaining content related to the target virtual object from the i-th round of modified text and historical parameters.

[0250] The Intent Recognizer 3024 analyzes the user's input and identifies the user's intention to modify the data. For example, based on the text to be modified in the i-th round and historical parameters, it determines the object to be modified in the i-th round. See Cases 1-3 above for details, which will not be repeated here.

[0251] Language model 3025 is used to transform the target virtual object and the modified text in the i-th round to obtain the target parameters. See A1-A3 or B1-B3 for details, which will not be repeated here.

[0252] Therefore, high-precision understanding of player intentions is achieved through deep learning and pre-trained models. A context manager ensures dialogue coherence and improves understanding accuracy. The module design considers real-time performance, enabling rapid response to user input. A microservice architecture facilitates module expansion and maintenance. Fine-tuning of pre-trained models adapts to the specific contexts of different games. Automated conversion from natural language descriptions to numerical parameters improves efficiency. User-customizable characters based on personal preferences enhance the personalization of the gaming experience.

[0253] (3) Optimization module 303 is used to modify the target parameters, obtain the adaptation value, and send the adaptation value to the colonel module 304.

[0254] Among them, compared to target parameters or standard parameters that are closer to the player's modification intentions, the adapted values ​​are more in line with the game settings.

[0255] See Figure 5 This figure is a schematic diagram of an optimization module provided in an embodiment of this application. Figure 5 In this module, optimization module 303 includes a parameterized representation generator 3031, a numerical optimizer 3032, a game rule adapter 3033, a user feedback interface 3034, and a prompt module 3035. These will be described in detail below.

[0256] The parameterization generator 3031 is used to convert target parameters into standard parameters. Compared to target parameters, which include textual descriptions and are easier to understand, standard parameters do not include textual descriptions and are easier for machines to understand.

[0257] The numerical optimizer 3032 is used to optimize standard parameters to improve the performance and effectiveness of virtual objects in the game.

[0258] Game rule adapter 3033 is used to convert standard parameters into updated target parameters. See D1-D5 for related details, the difference being that D1-D5 convert target parameters into updated target parameters, while game rule adapter 3033 can treat standard parameters as target parameters to obtain updated target parameters.

[0259] In addition, target parameters, standard parameters, or updated target parameters can be stored in a data warehouse for later retrieval.

[0260] User feedback interface 3034 is used to receive player feedback, such as confirmation of adaptive attribute values ​​and feedback on the performance of virtual objects, to guide further optimization.

[0261] The prompt module 3035 is used to determine the direction description of the (i+1)th modification based on the modification result parameter if the first object to be modified and the second object to be modified are the same, and the difference between the target parameter and the modification result parameter is greater than or equal to the difference threshold, and to display the direction description of the (i+1)th modification. See C1-C3 for details, which will not be elaborated here.

[0262] In addition, adaptation verification can be performed before sending the updated target parameters to the effective module 304 to ensure that the updated target parameters pass performance tests and conform to game settings.

[0263] Therefore, through numerical optimization, character performance in the game is significantly improved. Custom numerical values ​​are ensured to strictly adhere to game rules to avoid disrupting game balance. Numerical adjustments are made based on user feedback to meet player expectations and needs. The module can adapt to various game types and diverse user requirements. The module design considers scalability, facilitating the addition of new parameters and functions in the future. The modular design makes future expansion and maintenance easier.

[0264] (4) Activation module 304 is used to update the target virtual object based on the updated target parameters to obtain the modified target virtual object.

[0265] See Figure 6 This figure is a schematic diagram of an effective module provided in an embodiment of this application. Figure 6 The system includes a numerical application unit 3041, an effect display unit 3042, a user feedback interface 3043, and a data synchronizer 3044, which will be described below.

[0266] The numerical applicator 3041 is used to apply the updated target parameters to the corresponding target virtual object, ensuring that the updated target parameters are updated and reflected in the game in real time.

[0267] Effects Displayer 3042 is used to transform updated target parameters into in-game effects that players can perceive, such as character appearance, actions, and skill performance.

[0268] User feedback interface 3043 is used to provide a channel for user feedback and interaction, allowing players to evaluate and provide suggestions on the updated target parameters.

[0269] The data synchronizer 3044 is used to ensure the consistency and synchronization of numerical data between the game server and the client.

[0270] This allows players to instantly see the effects of their custom settings in the game. Real-time monitoring also ensures that updated target parameters do not negatively impact game performance. It provides a highly personalized gaming experience, meeting players' customization needs. It improves the stability of updated target parameters in the game, avoiding potential problems. The module can adapt to different game types and diverse numerical settings.

[0271] In response to the virtual object modification method described above, this application also provides a corresponding virtual object modification device so that the above virtual object modification method can be applied and implemented in practice.

[0272] See Figure 7 This figure is a schematic diagram of the structure of a virtual object modification device provided in an embodiment of this application. Figure 7 As shown, the virtual object modification device 700 includes: an acquisition unit 701, a determination unit 702, and a modification unit 703;

[0273] The acquisition unit 701 is used to acquire the i-th round of modified text from the target account, wherein the i-th round of modified text is used to describe the content corresponding to the i-th modification, and i is an integer greater than 1;

[0274] The acquisition unit 701 is further configured to acquire historical parameters based on the (i-1)th round of modified text, wherein the (i-1)th round of modified text comes from the target account, and the historical parameters are parameters used to describe the content corresponding to the (i-1)th modification indicated by the (i-1)th round of modified text;

[0275] The determining unit 702 is used to determine a target virtual object and target parameters based on the i-th round of modified text and the historical parameters, wherein the target virtual object is the virtual object corresponding to the i-th modification, and the target parameters are parameters used to describe the modified target virtual object;

[0276] The modification unit 703 is used to update the target virtual object based on the target parameters to obtain the modified target virtual object.

[0277] As can be seen from the above technical solution, the embodiments of this application provide a method for modifying virtual objects. Players do not need to directly modify parameters, but instead modify the corresponding content through natural language description, and then modify the natural language after converting it into parameters. This is not only convenient and quick, reducing the difficulty of modification, but also improving the flexibility and personalization of players' customized virtual objects. Moreover, in order to improve the effect of modification, multiple modifications can be performed through multiple rounds of natural language description. Taking the i-th modification as an example, since the modification intentions corresponding to multiple modifications may be consistent, the i-th round of modification text may not fully describe the modification intention. Therefore, in order to clarify the modification intention, after obtaining the i-th round of modification text sent through the target account, the historical parameters corresponding to the (i-1)-th round of modification text from the target account are obtained. These historical parameters can describe the modification intention corresponding to the (i-1)-th modification. Thus, based on the i-th round of modification text and the historical parameters, the target virtual object corresponding to the i-th modification and the target parameters used to describe the modified target virtual object are determined. Then, the target virtual object is updated based on the target parameters to realize the modification of the target virtual object.

[0278] Therefore, compared to determining the modification intent of the i-th modification by combining the i-1 round of modification text, the historical parameters corresponding to the i-1 round of modification text occupy less storage space, reducing storage pressure. Moreover, for the normal operation of the target virtual object, the historical parameters are content that should be stored. That is, compared to storing the i-1 round of modification text, the historical parameters are stored for a longer period of time. Thus, even if there is a long time difference between the i-th modification and the i-1th modification, the historical parameters used to describe the modification intent corresponding to the i-1th modification can still be obtained. Therefore, no matter how long the interval between two adjacent modifications is, the modification intent can be further clarified through contextual relationships, improving the accuracy of modification intent recognition, improving the modification effect, and thus improving the game's interaction rate and user retention rate.

[0279] As one possible implementation, the determining unit 702 is specifically used for:

[0280] The target virtual object is determined based on the i-th round of modified text and the historical parameters;

[0281] Based on the target virtual object and the i-th round of modified text, the target parameters are obtained by conversion using the language model;

[0282] The device further includes a training unit for:

[0283] Multiple modified text samples are obtained, each of which includes a label. The label is a parameter corresponding to the virtual object described by the modified text sample. The multiple modified text samples are used to describe multiple types of virtual objects, and there are multiple virtual objects of each type.

[0284] Based on the modified text sample and the virtual object described by the modified text sample, the prediction parameters are obtained by conversion through the initial language model;

[0285] Based on the difference between the prediction parameters and the labels of the modified text samples, the model parameters of the initial language model are adjusted to obtain the language model trained by the multiple modified text samples.

[0286] As one possible implementation, the determining unit 702 is specifically used for:

[0287] The target virtual object is determined based on the i-th round of modified text and the historical parameters;

[0288] The language model corresponding to the target virtual object is obtained from multiple language models, and different language models are applicable to different types of virtual objects;

[0289] Based on the target virtual object and the i-th round of modified text, the target parameters are obtained by conversion using the language model corresponding to the target virtual object;

[0290] The device further includes a training unit for:

[0291] Multiple modified text samples are obtained to describe the target virtual object. Each modified text sample includes a label, and the label is a parameter corresponding to the target virtual object described by the modified text sample. Different modified text samples correspond to different parameters.

[0292] Based on the modified text sample and the target virtual object, the prediction parameters are obtained by conversion using the initial language model of the target virtual object;

[0293] Based on the difference between the prediction parameters and the labels of the modified text samples, the model parameters of the initial language model of the target virtual object are adjusted to obtain the language model of the target virtual object trained by the multiple modified text samples.

[0294] As one possible implementation, the determining unit 702 is specifically used for:

[0295] If the first pending modification object is obtained by identifying the modified text in the i-th round, then the second pending modification object described by the historical parameters is obtained; if the first pending modification object and the second pending modification object are the same, then the first pending modification object or the second pending modification object is determined as the target virtual object; if the first pending modification object and the second pending modification object are not the same, then the first pending modification object is determined as the target virtual object.

[0296] If the first pending modification object is not obtained by identifying the modified text in the i-th round, then the second pending modification object described by the historical parameters is obtained, and the second pending modification object is determined as the target virtual object.

[0297] As one possible implementation, the determining unit 702 is specifically used for:

[0298] If the first pending modification object and the second pending modification object are different, then the target old version parameters used to describe the first pending modification object are obtained from the old version parameters obtained from the historical modification. The first pending modification object is a virtual object obtained based on the i-th round of modification text, and the second pending modification object is a virtual object obtained based on the (i-1)-th round of modification text. The historical modification refers to the modification made by the target account before the (i-1)-th modification.

[0299] Based on the i-th round of modified text and the target old version parameters, the target virtual object and the target parameters are determined.

[0300] As one possible implementation, the determining unit 702 is specifically used for:

[0301] Obtain the i-th modification time used to identify the i-th modification, and the historical modification time used to identify each of the historical modifications;

[0302] The target old version parameter is determined based on the time difference between the i-th modification time and the historical modification time. The time difference between the historical modification time corresponding to the target old version parameter and the i-th modification time is less than a time threshold, and the virtual object described by the target old version parameter is the first object to be modified.

[0303] As one possible implementation, the determining unit 702 is specifically used for:

[0304] If the first pending modification object and the second pending modification object are the same, then the modification result parameter is determined according to the i-th round of modification text and the historical parameters. The first pending modification object is a virtual object obtained based on the i-th round of modification text, and the second pending modification object is a virtual object obtained based on the (i-1)-th round of modification text. The modification result parameter is a parameter used to characterize the prediction of the modification target of the first pending modification object based on the target account.

[0305] If the difference between the target parameter and the modified result parameter is greater than or equal to the difference threshold, then the direction of the (i+1)th modification is determined based on the modified result parameter.

[0306] If the difference between the target parameter and the modified result parameter is less than the difference threshold, then the modification unit 703 is specifically used for:

[0307] The target virtual object is updated based on the modified result parameters to obtain the modified target virtual object.

[0308] As one possible implementation, if the target parameter includes multiple attribute pairs, and each attribute pair includes an attribute and its corresponding attribute value, then the modification unit 703 is specifically used for:

[0309] If there is a balance relationship between the first attribute and the second attribute, then the primary attribute and the secondary attribute are determined from the first attribute and the second attribute according to the i-th round of modified text and the historical parameters. The first attribute and the second attribute belong to attributes in different attribute pairs. In the balance relationship, a change in the attribute value of the primary attribute will cause a change in the attribute value of the secondary attribute.

[0310] Based on the primary attribute and the balance relationship, determine the range of suitable values ​​for the secondary attribute;

[0311] If the attribute value of the subordinate attribute exceeds the adaptation value range, the attribute value of the subordinate attribute is adjusted to obtain the adjusted attribute value of the subordinate attribute. The adjusted attribute value of the subordinate attribute and the attribute value of the main attribute satisfy the balance relationship.

[0312] Based on the adjusted attribute values ​​of the subordinate attributes, the target parameters are updated to obtain the updated target parameters;

[0313] The target virtual object is updated according to the updated target parameters to obtain the modified target virtual object.

[0314] As one possible implementation, if the balance relationship is a correlation trend curve, then the modification unit 703 is specifically used for:

[0315] Based on the correlation trend curve and the attribute value of the main attribute, determine the appropriate attribute value of the secondary attribute;

[0316] In response to obtaining a confirmation operation for the adapted attribute value, the adapted attribute value is determined as the adjusted attribute value of the subordinate attribute.

[0317] As one possible implementation, if there are multiple target virtual objects, then the modification unit 703 is specifically used for:

[0318] Obtain environmental data to describe the virtual environment, where the virtual environment is the environment in which the target virtual object is located;

[0319] Based on the environmental data, multiple target virtual objects are divided into a set of dominant virtual objects and a set of non-dominant virtual objects. The degree of integration between the dominant virtual objects in the set of dominant virtual objects and the environmental data is greater than or equal to a degree of integration threshold, and the degree of integration between the non-dominant virtual objects in the set of non-dominant virtual objects and the environmental data is less than the degree of integration threshold.

[0320] Obtain a first adjustment coefficient and a second adjustment coefficient, wherein the first adjustment coefficient is greater than the second adjustment coefficient;

[0321] The target parameters are adjusted according to the first adjustment coefficient to obtain the first target parameters. The advantageous virtual objects in the advantageous virtual object set are adjusted according to the first target parameters to obtain the adjusted advantageous virtual object set.

[0322] The target parameter is adjusted according to the second adjustment coefficient to obtain the second target parameter. The non-dominant virtual objects in the non-dominant virtual object set are then adjusted according to the second target parameter to obtain the adjusted non-dominant virtual object set.

[0323] The modified target virtual object is obtained based on the adjusted set of dominant virtual objects and the adjusted set of non-dominant virtual objects.

[0324] As one possible implementation, if the target parameter includes attribute pairs, and the attribute pairs include attributes described by text and attribute values ​​corresponding to the attributes, then the modification unit 703 is specifically used for:

[0325] Retrieve the correspondence between attributes described by text and attributes described by parameters;

[0326] Based on the correspondence, the attributes described by text in the target parameters are converted into attributes described by parameters to obtain standard parameters;

[0327] The target virtual object is updated based on the standard parameters to obtain the modified target virtual object.

[0328] This application also provides a computer device, which can be a server or a terminal device. The computer device provided in this application will be described below from a hardware implementation perspective. Figure 8 The diagram shown is a schematic of the server's structure. Figure 9 The diagram shown is a structural schematic of the terminal device.

[0329] See Figure 8 This figure is a schematic diagram of a server structure provided in an embodiment of this application. The server 1400 can vary considerably due to different configurations or performance. It may include one or more processors 1422, such as a central processing unit (CPU), memory 1432, and one or more application programs 1442 or data storage media 1430 (e.g., one or more mass storage devices). The memory 1432 and storage media 1430 can be temporary or persistent storage. The program stored in the storage media 1430 may include one or more modules (not shown in the figure), each module may include a series of instruction operations on the server. Furthermore, the processor 1422 may be configured to communicate with the storage media 1430 and execute the series of instruction operations in the storage media 1430 on the server 1400.

[0330] Server 1400 may also include one or more power supplies 1426, one or more wired or wireless network interfaces 1450, one or more input / output interfaces 1458, and / or one or more operating systems 1441, such as Windows Server. TM Mac OS X TM Unix TM Linux TM FreeBSD TM etc.

[0331] The steps performed by the server in the above embodiments can be based on this Figure 8 The server structure shown.

[0332] The processor 1422 is used to perform the following steps:

[0333] Obtain the i-th round of modified text from the target account, where the i-th round of modified text describes the content corresponding to the i-th modification, and i is an integer greater than 1;

[0334] Obtain historical parameters based on the (i-1)th round of modified text, where the (i-1)th round of modified text comes from the target account, and the historical parameters are parameters used to describe the content corresponding to the (i-1)th modification indicated by the (i-1)th round of modified text;

[0335] Based on the i-th round of modified text and the historical parameters, a target virtual object and target parameters are determined, wherein the target virtual object is the virtual object corresponding to the i-th modification, and the target parameters are parameters used to describe the modified target virtual object;

[0336] The target virtual object is updated based on the target parameters to obtain the modified target virtual object.

[0337] Optionally, the processor 1422 may also execute method steps of any specific implementation of the virtual object modification method in the embodiments of this application.

[0338] See Figure 9 This figure is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. The description will be based on a smartphone as an example. Figure 9 The diagram shown is a partial structural block diagram of the smartphone, which includes: a radio frequency (RF) circuit 1510, a memory 1520, an input unit 1530, a display unit 1540, a sensor 1550, an audio circuit 1560, a Wi-Fi module 1570, a processor 1580, and a power supply 1590, among other components. Those skilled in the art will understand that... Figure 9 The smartphone structure shown does not constitute a limitation on smartphones and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0339] The following is combined Figure 9 A detailed introduction to the various components of a smartphone:

[0340] The RF circuit 1510 can be used to receive and transmit signals during information transmission or calls. In particular, it receives downlink information from the base station and processes it with the processor 1580; in addition, it transmits uplink data to the base station.

[0341] The memory 1520 can be used to store software programs and modules, and the processor 1580 runs the software programs and modules stored in the memory 1520 to realize various functions and data processing of the smartphone.

[0342] Input unit 1530 can be used to receive input numeric or character information and generate key signal inputs related to user settings and function control of the smartphone. Specifically, input unit 1530 may include touch panel 1531 and other input devices 1532. Touch panel 1531, also known as a touch screen, can collect touch operations on or near the user and drive corresponding connected devices according to a pre-set program. In addition to touch panel 1531, input unit 1530 may also include other input devices 1532. Specifically, other input devices 1532 may include, but are not limited to, one or more of the following: physical keyboard, function keys (such as volume control buttons, power buttons, etc.), trackball, mouse, joystick, etc.

[0343] The display unit 1540 can be used to display information input by the user or information provided to the user, as well as various menus of the smartphone. The display unit 1540 may include a display panel 1541, which may optionally be configured as a liquid crystal display (LCD), an organic light-emitting diode (OLED), or the like.

[0344] Smartphones may also include at least one sensor 1550, such as a light sensor, a motion sensor, and other sensors. Other sensors that smartphones may also be equipped with, such as gyroscopes, barometers, hygrometers, thermometers, and infrared sensors, will not be detailed here.

[0345] Audio circuit 1560, speaker 1561, and microphone 1562 provide an audio interface between the user and the smartphone. Audio circuit 1560 converts received audio data into electrical signals and transmits them to speaker 1561, where speaker 1561 converts them into sound signals for output. On the other hand, microphone 1562 converts collected sound signals into electrical signals, which are received by audio circuit 1560, converted into audio data, and then processed by processor 1580 before being transmitted via RF circuit 1510 to, for example, another smartphone, or the audio data can be output to memory 1520 for further processing.

[0346] The processor 1580 is the control center of the smartphone, connecting various parts of the smartphone through various interfaces and lines. It performs various functions and processes data by running or executing software programs and / or modules stored in the memory 1520, and by calling data stored in the memory 1520. Optionally, the processor 1580 may include one or more processing units.

[0347] The smartphone also includes a power supply 1590 (such as a battery) that supplies power to various components. Preferably, the power supply can be logically connected to the processor 1580 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system.

[0348] Although not shown, smartphones may also include a camera, Bluetooth module, etc., which will not be described in detail here.

[0349] In this embodiment of the application, the memory 1520 included in the smartphone can store computer programs and transmit the computer programs to the processor.

[0350] The processor 1580 included in the smartphone can execute the virtual object modification method provided in the above embodiments according to the instructions in the computer program.

[0351] This application also provides a computer-readable storage medium for storing a computer program for executing the virtual object modification method provided in the above embodiments.

[0352] On the other hand, embodiments of this application provide a computer program product including a computer program, which, when run on a computer device, causes the computer device to perform the virtual object modification method provided in various optional implementations of the above aspects.

[0353] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium can be at least one of the following media: read-only memory (ROM), RAM, magnetic disk or optical disk, and other media that can store computer programs.

[0354] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0355] In this application embodiment, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0356] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing 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 be referred to the description of the method embodiments. The device and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the solution in this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0357] The above description is merely one specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Based on the implementation methods provided in the above aspects, this application can also be further combined to provide more implementation methods. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for modifying a virtual object, characterized in that, The method includes: Obtain the i-th round of modified text from the target account, where the i-th round of modified text describes the content corresponding to the i-th modification, and i is an integer greater than 1; Obtain historical parameters based on the (i-1)th round of modified text, where the (i-1)th round of modified text comes from the target account, and the historical parameters are parameters used to describe the content corresponding to the (i-1)th modification indicated by the (i-1)th round of modified text; Based on the i-th round of modified text and the historical parameters, a target virtual object and target parameters are determined, wherein the target virtual object is the virtual object corresponding to the i-th modification, and the target parameters are parameters used to describe the modified target virtual object; The target virtual object is updated based on the target parameters to obtain the modified target virtual object.

2. The method according to claim 1, characterized in that, The step of determining the target virtual object and target parameters based on the i-th round of modified text and the historical parameters includes: The target virtual object is determined based on the i-th round of modified text and the historical parameters; Based on the target virtual object and the i-th round of modified text, the target parameters are obtained by conversion using the language model; The language model is trained as follows: Multiple modified text samples are obtained, each of which includes a label. The label is a parameter corresponding to the virtual object described by the modified text sample. The multiple modified text samples are used to describe multiple types of virtual objects, and there are multiple virtual objects of each type. Based on the modified text sample and the virtual object described by the modified text sample, the prediction parameters are obtained by conversion through the initial language model; Based on the difference between the prediction parameters and the labels of the modified text samples, the model parameters of the initial language model are adjusted to obtain the language model trained by the multiple modified text samples.

3. The method according to claim 1, characterized in that, The step of determining the target virtual object and target parameters based on the i-th round of modified text and the historical parameters includes: The target virtual object is determined based on the i-th round of modified text and the historical parameters; The language model corresponding to the target virtual object is obtained from multiple language models, and different language models are applicable to different types of virtual objects; Based on the target virtual object and the i-th round of modified text, the target parameters are obtained by conversion using the language model corresponding to the target virtual object; The language model of the target virtual object is trained as follows: Multiple modified text samples are obtained to describe the target virtual object. Each modified text sample includes a label, and the label is a parameter corresponding to the target virtual object described by the modified text sample. Different modified text samples correspond to different parameters. Based on the modified text sample and the target virtual object, the prediction parameters are obtained by conversion using the initial language model of the target virtual object; Based on the difference between the prediction parameters and the labels of the modified text samples, the model parameters of the initial language model of the target virtual object are adjusted to obtain the language model of the target virtual object trained by the multiple modified text samples.

4. The method according to claim 2 or 3, characterized in that, The step of determining the target virtual object based on the i-th round of modified text and the historical parameters includes: If the first pending modification object is obtained by identifying the modified text in the i-th round, then the second pending modification object described by the historical parameters is obtained; if the first pending modification object and the second pending modification object are the same, then the first pending modification object or the second pending modification object is determined as the target virtual object; if the first pending modification object and the second pending modification object are not the same, then the first pending modification object is determined as the target virtual object. If the first pending modification object is not obtained by identifying the modified text in the i-th round, then the second pending modification object described by the historical parameters is obtained, and the second pending modification object is determined as the target virtual object.

5. The method according to claim 1, characterized in that, The step of determining the target virtual object and target parameters based on the i-th round of modified text and the historical parameters includes: If the first pending modification object and the second pending modification object are different, then the target old version parameters used to describe the first pending modification object are obtained from the old version parameters obtained from the historical modification. The first pending modification object is a virtual object obtained based on the i-th round of modification text, and the second pending modification object is a virtual object obtained based on the (i-1)-th round of modification text. The historical modification refers to the modification made by the target account before the (i-1)-th modification. Based on the i-th round of modified text and the target old version parameters, the target virtual object and the target parameters are determined.

6. The method according to claim 5, characterized in that, The step of obtaining target old version parameters for describing the first object to be modified from the old version parameters obtained from historical modifications includes: Obtain the i-th modification time used to identify the i-th modification, and the historical modification time used to identify each of the historical modifications; The target old version parameter is determined based on the time difference between the i-th modification time and the historical modification time. The time difference between the historical modification time corresponding to the target old version parameter and the i-th modification time is less than a time threshold, and the virtual object described by the target old version parameter is the first object to be modified.

7. The method according to claim 1, characterized in that, The method further includes: If the first pending modification object and the second pending modification object are the same, then the modification result parameter is determined according to the i-th round of modification text and the historical parameters. The first pending modification object is a virtual object obtained based on the i-th round of modification text, and the second pending modification object is a virtual object obtained based on the (i-1)-th round of modification text. The modification result parameter is a parameter used to characterize the prediction of the modification target of the first pending modification object based on the target account. If the difference between the target parameter and the modified result parameter is greater than or equal to the difference threshold, then the direction of the (i+1)th modification is determined based on the modified result parameter. If the difference between the target parameter and the modified result parameter is less than the difference threshold, then updating the target virtual object based on the target parameter to obtain the modified target virtual object includes: The target virtual object is updated based on the modified result parameters to obtain the modified target virtual object.

8. The method according to claim 1, characterized in that, If the target parameters include multiple attribute pairs, and each attribute pair includes an attribute and its corresponding attribute value, then updating the target virtual object based on the target parameters to obtain the modified target virtual object includes: If there is a balance relationship between the first attribute and the second attribute, then the primary attribute and the secondary attribute are determined from the first attribute and the second attribute according to the i-th round of modified text and the historical parameters. The first attribute and the second attribute belong to attributes in different attribute pairs. In the balance relationship, a change in the attribute value of the primary attribute will cause a change in the attribute value of the secondary attribute. Based on the primary attribute and the balance relationship, determine the range of suitable values ​​for the secondary attribute; If the attribute value of the subordinate attribute exceeds the adaptation value range, the attribute value of the subordinate attribute is adjusted to obtain the adjusted attribute value of the subordinate attribute. The adjusted attribute value of the subordinate attribute and the attribute value of the main attribute satisfy the balance relationship. Based on the adjusted attribute values ​​of the subordinate attributes, the target parameters are updated to obtain the updated target parameters; The target virtual object is updated according to the updated target parameters to obtain the modified target virtual object.

9. The method according to claim 8, characterized in that, If the balance relationship is a correlation trend curve, then adjusting the attribute value of the dependent attribute to obtain the adjusted attribute value includes: Based on the correlation trend curve and the attribute value of the main attribute, determine the appropriate attribute value of the secondary attribute; In response to obtaining a confirmation operation for the adapted attribute value, the adapted attribute value is determined as the adjusted attribute value of the subordinate attribute.

10. The method according to claim 1, characterized in that, If there are multiple target virtual objects, then updating the target virtual objects based on the target parameters to obtain the modified target virtual objects includes: Obtain environmental data to describe the virtual environment, where the virtual environment is the environment in which the target virtual object is located; Based on the environmental data, multiple target virtual objects are divided into a set of dominant virtual objects and a set of non-dominant virtual objects. The degree of integration between the dominant virtual objects in the set of dominant virtual objects and the environmental data is greater than or equal to a degree of integration threshold, and the degree of integration between the non-dominant virtual objects in the set of non-dominant virtual objects and the environmental data is less than the degree of integration threshold. Obtain a first adjustment coefficient and a second adjustment coefficient, wherein the first adjustment coefficient is greater than the second adjustment coefficient; The target parameters are adjusted according to the first adjustment coefficient to obtain the first target parameters. The advantageous virtual objects in the advantageous virtual object set are adjusted according to the first target parameters to obtain the adjusted advantageous virtual object set. The target parameter is adjusted according to the second adjustment coefficient to obtain the second target parameter. The non-dominant virtual objects in the non-dominant virtual object set are then adjusted according to the second target parameter to obtain the adjusted non-dominant virtual object set. The modified target virtual object is obtained based on the adjusted set of dominant virtual objects and the adjusted set of non-dominant virtual objects.

11. The method according to claim 1, characterized in that, If the target parameter includes attribute pairs, and the attribute pairs include attributes described by text and attribute values ​​corresponding to the attributes, then updating the target virtual object based on the target parameter to obtain the modified target virtual object includes: Retrieve the correspondence between attributes described by text and attributes described by parameters; Based on the correspondence, the attributes described by text in the target parameters are converted into attributes described by parameters to obtain standard parameters; The target virtual object is updated based on the standard parameters to obtain the modified target virtual object.

12. A virtual object modification device, characterized in that, The device includes: an acquisition unit, a determination unit, and a modification unit; The acquisition unit is used to acquire the i-th round of modified text from the target account, wherein the i-th round of modified text is used to describe the content corresponding to the i-th modification, and i is an integer greater than 1; The acquisition unit is further configured to acquire historical parameters based on the (i-1)th round of modified text, wherein the (i-1)th round of modified text comes from the target account, and the historical parameters are parameters used to describe the content corresponding to the (i-1)th modification indicated by the (i-1)th round of modified text; The determining unit is configured to determine a target virtual object and target parameters based on the i-th round of modified text and the historical parameters, wherein the target virtual object is the virtual object corresponding to the i-th modification, and the target parameters are parameters used to describe the modified target virtual object; The modification unit is used to update the target virtual object based on the target parameters to obtain the modified target virtual object.

13. A computer device, characterized in that, The computer device includes a processor and memory: The memory is used to store computer programs and to transfer the computer programs to the processor; The processor is configured to perform the method according to any one of claims 1-11 according to the computer program.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program for performing the method according to any one of claims 1-11.

15. A computer program product comprising a computer program, characterized in that, When it is run on a computer device, it causes the computer device to perform the method described in any one of claims 1-11.

Images